REVIEWS

NMDA RECEPTORS, PLACE CELLS AND

HIPPOCAMPAL SPATIAL MEMORY
Kazu Nakazawa* ‡§, Thomas J. McHugh*‡, Matthew A. Wilson‡ and Susumu Tonegawa*‡
N-methyl-D-aspartate receptors (NMDARs) in the rodent hippocampus have been shown to be
essential for spatial learning and memory, and for the induction of long-term synaptic plasticity at
various hippocampal synapses. In this review, we examine the evidence concerning the role of
NMDARs in hippocampal memory processes, with an emphasis on the function of NMDARs in
area CA1 of the hippocampus in memory acquisition, and the unique role of NMDARs in area
CA3 in the rapid acquisition and associative retrieval of spatial information. Finally, we discuss the
data that have emerged from in vivo hippocampal recording studies that indicate that the activity
of hippocampal place cells during behaviour is an expression of a memory trace.

*Howard Hughes Medical
Institute and
‡
The Picower Center for
Learning and Memory,
RIKEN-MIT Neuroscience
Research Center, Center for
Cancer Research, and
Departments of Biology and
Brain and Cognitive
Sciences, Massachusetts
Institute of Technology,
Cambridge, Massachusetts
02139, USA.
§
National Institute of
Mental Health, National
Institutes of Health, 9000
Rockville Pike, Bethesda,
Maryland 20892, USA.
Correspondence to S.T.
e-mail: tonegawa@mit.edu
doi:10.1038/nrn1385
Scoville and Milner’s study in the 1950s of patient H.M.,
who had undergone a bilateral resection of the hippo-
campus and associated cortical areas, indicated that these
parts of the brain, collectively called the medial temporal
lobe (MTL), are crucial in the formation of declarative
memory (memory of facts and events)1. Subsequent
studies on H.M. and other patients, as well as studies on
animal models, have enriched our knowledge about the
roles of the MTL and its components in declarative
memory2–5. It has become clear that the hippocampus
has an essential role in, among other types of memory,
spatial memory — a type of declarative memory that
is concerned with spatial locations6,7. In 1971, O’Keefe
and Dostrovsky discovered ‘place cells’8 (BOX 1), showing
that space can be encoded in the firing pattern of the
hippocampus.
Concurrent with these studies, Bliss and Lømo
discovered that high-frequency stimulation of the
hippocampal input fibres can result in long-lasting
enhancement of transmission efficacy at downstream
synapses9. This discovery of long-term potentiation
(LTP) provided the first experimental support for
Hebb’s theory on the neural representation of memory10.
Hippocampal LTP has subsequently been subjected
to extensive study as a candidate mechanism for
learning and memory. The induction of LTP in area
CA1 of the hippocampus is blocked by D(–)-2-amino-5-
phosphonovaleric acid (AP5), an antagonist of the
NMDA subset of glutamate receptors11. The NMDAR
possesses a voltage-dependent magnesium block,
high calcium permeability and slow activation and
deactivation kinetics12–15. Consequently, the NMDAR
can be opened by glutamate only when the postsynaptic
neuron is depolarized, thereby allowing the receptor to
function as a detector and integrator of coincident
activity at the synapse. The finding that the induction
of hippocampal LTP depends on the activation of
NMDARs further strengthened the link between LTP
and Hebb’s synaptic hypothesis for memory storage, in
which modifications of synaptic efficacy by coincident
input was the central theme. The demonstration that
infusion of AP5 into the ventricles of rats caused an
impairment in spatial learning indicated that NMDAR
activation might be crucial for this type of memory16.
Studies on the role of NMDARs in memory inten-
sified when the genes encoding the receptor were
cloned during the early 1990s17, allowing the NMDAR
subunits to be identified and characterized. Among the
seven identified subunits (NR1, NR2A–D, and NR3A
and B), the NR1 subunit is the only one that is indis-
pensable for the formation of a functional receptor; its
elimination would abolish all functional NMDARs in a
cell18,19. As AP5 inhibits LTP induction in hippocampal
slices and spatial memory in intact animals, it might
be expected that elimination of the NR1 gene would
lead to similar physiological and behavioural deficits.

NATURE REVIEWS | NEUROSCIENCE VOLUME 5 | MAY 2004 | 3 6 1

©2004 Nature Publishing Group

REVIEWS

Box 1 | Place cells

In 1971, O’Keefe and Dostrovsky a
Cell 1
found that single hippocampal
neurons increased their firing rate Cell 2
whenever a rat traversed a particular Cell 3
region of a chamber 8. When they Cell 4
recorded extracellular action
potentials from the hippocampus of
freely moving rats, some hippocampal
pyramidal cells demonstrated firing
b
patterns that seemed to depend on the
animal’s location in the environment.
When the rat left the ‘place’ that was
encoded by a given cell, the cell fell
almost silent. In an open field, the
firing rate was also independent of the
direction in which the animal entered
the area and the direction that it was
facing. The firing of each cell seemed to
indicate a specific location in the
environment of the rat; so these cells
are called ‘place cells’.
Panel a represents the place-specific
firing properties of hippocampal
pyramidal cells as a rodent runs down a
linear track. Each cell fires only on a
specific region of the track. A common
method of representing these place
fields (right) is a firing rate map — a
top-down view of the environment
with areas of high firing rate coloured red and yellow and areas with no firing coloured blue. Panel b shows 80 firing rate
maps of cells simultaneously recorded from area CA1 of a rat exploring a square arena. Most cells are silent as the rat
forages, with only about 30% of the pyramidal cells active in this environment. The six cells that fire throughout the
environment are thought to be interneurons125.
Since 1971, neuroscientists have conducted hundreds of studies to characterize the properties and dependencies of
place-cell activity153–156. Studies using freely behaving rats have shown that pyramidal cells show stable, long-lasting,
environmentally specific place fields, with between 30% and 50% of the CA1 cell layer showing place-specific activity in
any given environment125,157,158. When an animal is introduced to a novel environment, these place cells form rapidly,
usually within five minutes, and are maintained robustly125. The relative locations of these place-receptive fields change
in different environments, with no apparent topographical relationship to cell position, so new place fields must be
learned in each environment159.
More recent studies have shown that any given place is encoded not by the activity of single neurons, but instead by a
population of simultaneously active cells125. Parameters such as the coefficient of variance, a measure of temporal
correlation, allow the spatial coding properties of an ensemble of cells to be assessed. For example, both pairs and sequences
of cells that were active during behaviour, owing to close proximity of their place fields, tend to be reactivated in a
coordinated manner during periods of slow-wave and rapid eye movement (REM) sleep following behaviour124,127,128,132,133.
This increase in correlated activity during sleep might reflect cellular and network learning mechanisms.
Place cells have been best characterized in area CA1 of the rat hippocampus, but they have also been studied in the
other hippocampal subfields and in the entorhinal cortex74,75. The recording and analysis of fields in the genetically
modified mouse has also been applied to understanding the hippocampal code for space. Panel b reproduced, with
permission, from REF. 125 © (1993) American Association for the Advancement of Science.

However, NR1-knockout mice do not survive for more
than a day after birth because of the role of this receptor
in the midbrain for breathing20,21. This illustrates the
need for spatial and temporal restriction of genetic
interventions for the effective study of memory and
other cognitive phenomena.
The role of the NMDAR in spatial representation and
spatial memory in the hippocampus must be considered
and studied in the context of hippocampal anatomy.
As in humans and non-human primates, the rodent
hippocampus can be divided into the dentate gyrus
(DG), area CA3 and area CA1 (FIG. 1). In the main excita-
tory pathway, the trisynaptic circuit, information flows
from layer II stellate cells of the entorhinal cortex (EC) to
granule cells in the DG though the perforant path, from
the granule cells to the CA3 pyramidal cells through the
mossy fibres, from the CA3 pyramidal cells to the CA1
pyramidal cells though the Schaffer collaterals, and finally
from the CA1 pyramidal cells to cells in the deep layers
of the EC. In addition, layer II stellate cells and layer III

362 | MAY 2004 | VOLUME 5 www.nature.com/reviews/neuro

©2004 Nature Publishing Group


Fornix
Sub
CA1 5/6
3
PHG
2 Neocortex
and
PR
EC
DG
CA3
Figure 1 | Connections in the corticohippocampal network. Sensory information from the
cerebral association neocortex (neocortex) arrives at the superficial layers of the entorhinal
cortex (EC) by way of the parahippocampal gyrus (PHG) and/or perirhinal cortex (PR). From the
EC, the information is routed into the hippocampus. The main subfields of the hippocampus and
the connections among them and with the superficial layers of the EC are described in the text.
CA1 sends axons to both subcortical areas and the deep layers of the EC, either directly or
through the subiculum (Sub). The corticohippocampal network is then completed, with the
information from CA1 being sent back to cerebral association cortex through the PHG and/or
PR. Thick lines below or to the right of cell bodies (triangles) represent dendrites. Numbers refer
to cortical layers. DG, dentate gyrus.
pyramidal cells of the EC send axons directly to area CA3
(through the perforant path) and area CA1 (through the
temporoammonic pathway), respectively 22. Furthermore,
CA3 pyramidal cells are interconnected by recurrent
collaterals that run both ipsilaterally and contralaterally.
Among the three types of excitatory input that CA3 pyra-
midal cells receive, these recurrent collateral inputs are the
most numerous (about 12,000 per pyramidal cell in
the rat), whereas the perforant path and mossy fibre
inputs provide about 4,000 and 50 inputs per cell, respec-
tively. Owing to the recurrent collaterals, the connectivity
among CA3 pyramidal cells is robust; a given cell is
directly connected with at least 2% of the other cells23,24.
A number of theoretical studies have proposed a
distinct mnemonic role for each of the hippocampal
subfields and inputs25–28. For instance, it has been postu-
lated that the recurrent network in CA3 is crucial for the
storage of associative memory and its recall by ‘pattern
completion’, whereas the DG is involved in the separa-
tion of similar memories (‘pattern separation’). It has
also been suggested that the CA1 network might be
instrumental in recognizing the novelty or familiarity of
an object or context29. However, it is only recently that
some of these hypotheses have been tested empirically.
Memory acquisition
Early support for the link among NMDAR activity,
hippocampal LTP and learning and memory came from
Morris and colleagues. They developed a hippocampus-
dependent behavioural task for rats using a circular
pool filled with opaque water in which an escape
platform was hidden at a fixed location below the surface
NATURE REVIEWS | NEUROSCIENCE
©2004 Nature Publishing Group
REVIEWS
(the hidden-platform version of the Morris water maze
(MWM) task 6,30). Initially, rats that are placed in the pool
find the escape platform by random navigation. How-
ever, over repeated training trials, they slowly learn
the location using objects outside the maze as cues and
eventually swim nearly straight to the platform. Blockade
of NMDARs by intraventricular infusion of AP5 resulted
in an impairment of spatial learning on the hidden-
platform MWM task, but not if the platform was
visible16. The estimated extracellular concentrations of
AP5 in the hippocampus that caused the spatial learning
impairment were comparable to those that blocked LTP
in vivo at synapses between the perforant path and
the DG31. Intrahippocampal injection of AP5 resulted in
similar behavioural deficits32. By contrast, intrahippo-
campal administration of an AMPA (α-amino-3-
hydroxy-5-methyl-4-isoxazole propionic acid)/kainate
receptor antagonist, 6-cyano-7-nitroquinoxaline-2,3-
dione (CNQX) or LY326325, resulted in a severe memory
deficit with either the hidden or the visible platform33.
These results indicated that hippocampal NMDAR
activity and NMDAR-dependent plasticity are crucial
for spatial learning. However, NMDAR antagonists
could have nonspecific effects on neural activity. For
instance, AP5 can cause a reduction in hippocampal
excitability34,35. Furthermore, intraventricular infusion
of AP5 often results in diffusion of the drug to areas
outside the hippocampus, which could impair both
sensory and motor function36,37. The possibility that
unintended effects of AP5 contributed to the observed
learning deficits could not be excluded.
A complementary approach was provided by geneti-
cally engineered mice, in which deletion of the NR1
gene was restricted to particular cell types using the phage
P1-derived, Cre/loxP recombination system38,39 (BOX 2).
Tsien et al. created a mouse strain in which the NR1 gene
was postnatally knocked out predominantly in CA1
pyramidal cells (referred to as CA1-NR1-knockout
mice)40. This mouse strain had apparently normal
growth and was fertile.
CA1-NR1-knockout mice showed severely impaired
spatial learning in the hidden-platform MWM, but
performed normally when the platform was visible40. A
possible explanation for this deficit came from in vivo
recordings of CA1 pyramidal cells41. Place cells that were
recorded while CA1-NR1-knockout mice traversed a lin-
ear track had altered single-cell and ensemble properties.
The individual fields were larger and less structured than
normal, and the coefficient of variance — a measure of
the temporal coordination of the firing of cells with over-
lapping place fields — was greatly reduced, indicating a
loss of coherent spatial representation at the ensemble
level. Furthermore, the in vitro induction of LTP was
specifically blocked at the Schaffer collateral (SC)–CA1
synapses. In these mice, NR1 deletion is delayed
until about 4 weeks after birth and is restricted to CA1
pyramidal cells until about 2 months of age42. So, it
is unlikely that the behavioural impairment is the result
of undetected developmental abnormalities, and
although a small percentage of the animals that were used
for behavioural and physiological characterization might
VOLUME 5 | MAY 2004 | 3 6 3
REVIEWS
FEAR CONDITIONING
A form of Pavlovian (classical)
conditioning in which the
animal learns that an innocuous
stimulus (for example, an
auditory tone — the
conditioned stimulus or CS),
reliably predicts the occurrence
of a noxious stimulus (for
example, foot shock — the
unconditioned stimulus or US)
following their repeated paired
presentation. As a result of this
procedure, presentation of the
CS alone elicits conditioned fear
responses previously associated
with the noxious stimulus only.
TRANSVERSE PATTERN LEARNING
A task in which animals must
encode overlapping
relationships between cues. A
typical stimulus set is A+B–;
B+C–; C+A–, where + signifies
which cue is rewarded in each
configuration.
364 | MAY 2004 | VOLUME 5
Box 2 | Conditional genetic manipulations
Conditional genetic techniques have been used in rodents a
to confine the manipulation of a gene to a particular tissue
or cell type and/or a desired time point, allowing spatial PαCamKll Cre loxP loxP
and temporal control of a gene of interest in vivo. The
Cre/loxP system is the most widely used technique for NR1 exons
Cre
manipulating the mouse genome160,161. Cre is a site-specific
Mouse A Mouse B
DNA recombinase derived from the P1 bacteriophage that floxed NR1
CA1-specific Cre
recognizes 34 base-pair sequences termed loxP sites. Cre
catalyses the deletion of a segment of DNA that is flanked CA1 pyramidal cells All other cells (Cre absent)
(Cre present)
by a pair of these loxP sites (floxed), resulting in a
‘knockout’ of the gene of interest (a). To create mice in
which a gene knockout is restricted to a brain subregion,
two transgenic mouse lines are created and intercrossed:
one line is composed of mice in which a pair of loxP sites
are introduced by homologous recombination into the
b
gene to be knocked out and the other line comprises
transgenic mice in which the expression of the Cre
recombinase is driven by a tissue-specific or cell-type-
specific transcriptional promoter (a). When these are bred floxed NR1 floxed NR1/Cre
together, the restricted expression of the recombinase c
PαCamKll tTA tetO Pmin NR1
leads to a specific deletion of the floxed gene only in the
tissue of interest. The use of promoters that are active only
in the adult minimizes the developmental and tTA
compensatory effects that are often seen in conventional Mouse A Mouse B
knockout mice162–164. Panel b shows in situ hybridization Forebrain-specific tTA tetO-NR1
using a probe that is specific for the NR1 transcript. In the Forebrain cells (tTA present) All other cells (tTA absent)
absence of the recombinase, the transcript is expressed
throughout the brain (left); however, Cre expression leads
to a CA1-specific deletion of the gene (arrow, right). + Doxycycline ( )
To endow temporal specificity to a genetic alteration,
there are various techniques for creating mice with
inducible expression of genes165. In the nervous system, the
most widely used inducible approach is the tetracycline-dependent regulatory system166. The tetracycline transactivator
(tTA), a fusion of the tet repressor protein of Escherichia coli and the carboxy (C)-terminal domain of the herpes simplex
virus VP16 protein, is used as a transcriptional ON/OFF switch to drive the expression of a gene of interest109. Like the
Cre/loxP system, the tTA system requires the use of two lines of transgenic mice — mice in which the expression of the
tTA protein is driven by a tissue or cell-type-specific promoter, and mice in which a gene of interest is placed downstream
of a tandem array of the tet operator (tetO), along with a minimal promoter (c). In the original and most frequently used
version of this system, the tTA protein binds to the tetO sequence and induces transcription. However, when the inducer
(tetracycline) is present, it binds to tTA and prevents it from binding to tetO, which halts transcription. Many
laboratories have now reported the successful application of the tTA system to neuroscience, observing that removal of
tetracycline from the animals’ food or water leads to robust expression of the exogenous transgenic protein in the
brain167–169. The application and combination of spatial restriction by the Cre/loxP system and of temporal control by the
tTA system has allowed the generation of more precise reagents for the study of systems neuroscience.
have harboured a more widespread deletion of the NR1 conditioning and TRANSVERSE PATTERN LEARNING43–45. However,
gene, most of the analysis was done with young mice in we focus on spatial memory in this review, to compare
which the knockout was CA1-specific. In addition, the and contrast the molecular genetic, behavioural, pharma-
lack of NR1 deletion in the neocortex alleviated the cological and electrophysiological data that support our
main concern that accompanied the AP5 administra- evaluation of NMDARs in hippocampal function.
tion experiments. The combination of pharmacological
and genetic studies has therefore provided strong evi- Pretraining spares spatial learning. Despite the evidence
dence that NMDAR activity and NMDAR-dependent outlined above, some subsequent studies called into
synaptic plasticity in the hippocampus are crucial question the role of NMDAR-dependent plasticity in
for spatial learning and for the proper formation and spatial learning. Most strikingly, NMDAR antagonists
coordination of CA1 place fields. The importance of failed to disrupt new spatial learning in the hidden-
CA1 NMDARs in the acquisition of hippocampus- platform MWM task in rats that had received spatial46
dependent memory was subsequently extended to or non-spatial47 pretraining in a different environment.
various non-spatial tasks, including recognition of The hippocampus was still needed for new spatial learn-
novel objects, contextual FEAR CONDITIONING, trace fear ing by the pretrained animals46. Pretraining also protects
www.nature.com/reviews/neuro
©2004 Nature Publishing Group
 REVIEWS

a 5 min 5 min 5 min Spatial delayed matching-to-place task. One clue came
T1 T2 T3 T4 from the effects of AP5 in the delayed matching-to-
place (DMP) version of the MWM50 (FIG. 2). In this task,
the location of the platform was altered daily in a
Day 1 pseudo-random fashion. On each day, four trials were
given with intertrial intervals of 20 min or 2 h. The
mean latencies to find the platform are relatively long
for the first trial of a day, because the animals are
unaware of the novel location of the platform, but for
untreated rats they are significantly shortened in the
second trial owing to the memory that was acquired
Day 2
earlier that day. However, infusion of AP5 into the
hippocampus substantially reduced the latency savings
on the second trial. Pretraining over 9 days with a novel
platform location each day did not alter the NMDAR-
dependency of this one-trial spatial learning. This is
in contrast to the results of the MWM task in which
pretraining absolved the need for NMDAR function in
Day n new spatial learning. What features of the two water
maze tasks resulted in this difference?
In the DMP task, the animals have presumably
familiarized themselves with both the task rules and the
b Day 1–4 Day 5–8 Day 13–16 environment by the end of the pretraining session.
100 100 100 To reduce the latency on the second trial of a test day,
80 the animal must depend on information about the
Escape latency (s)

80
Escape latency (s)
Escape latency (s)

80
platform location that it obtained during the first trial
60 60 60
of the day. The animals must distinguish the most
40 40 40 recent (today’s) platform location from those experi-
enced in the preceding days. These requirements are
20 20 20
reminiscent of those for human episodic memory51,52.
0 0 0
Although it has been forcefully argued by some that
1 2 3 4 1 2 3 4 1 2 3 4 episodic memory is beyond the capability of non-
Trials Trials Trials
human animals53, recent studies on food-caching and
Figure 2 | CA3-NR1-knockout mice are impaired in delayed matching-to-place task in recovery in scrub jays54 indicate that animals can mem-
water maze. a | Animals are given four trials (T1–T4) per day with a 5-min intertrial interval. During
a given day the hidden platform remains at a fixed location, but between days, the platform is
orize ‘what’, ‘where’ and ‘when’ information, a hallmark
moved to novel locations. b | Performance is averaged across 16-trial blocks during training (day of episodic memory55. Hippocampal NMDARs might
1–4, 5–8, 9–12 and 13–16). The escape latency is relatively long on trial 1 each day because the be needed for this episodic aspect of spatial learning56.
animal will have to conduct a random search for the novel platform location, but it is reduced In contrast to the DMP task, which demands rapid
gradually on trials 2, 3 and 4. Up to and including day 12, there were no differences in the escape acquisition of the memory for the novel platform
latencies between CA3-NR1-knockout mice and their control littermates, indicating that the location in one trial, the hidden-platform MWM
mutants are not impaired in the ‘rule learning’ phase of the task. On days 13–16, however, the
task allows slow acquisition of the memory of the
mutants exhibited latency deficits in trials 2–4. These results indicate that the mutants are
impaired in the rapid encoding of novel spatial information69. fixed platform location, and its performance can be
improved incrementally throughout multiple trials.
During the early stages of training, animals probably
learning on a STEP-DOWN INHIBITORY AVOIDANCE TASK in rats rely on single-trial memory, sometimes referred to as
that have received an intrahippocampal infusion of AP5 episodic information, for the performance improve-
(REF. 48). The NMDAR dispensability that is induced by ment as in the DMP task, but as the training advances
pretraining seems to reflect the complexity of the learning they might gradually form more trial-independent
associated with these behavioural tasks. In the hidden- spatial memory. We suggest that hippocampal NMDARs
platform MWM task, animals must learn various behav- are required in the early, trial-dependent phase of the
ioural strategies and rules, including swimming away task, but not in the later, trial-independent phase. When
from the side walls to find a platform, discovering the animals undergo extensive pretraining on this task,
hidden platform, and climbing and staying on the plat- even in a distinct environment, it might minimize
form, before they can learn the location of the platform the trial-dependent phase of new spatial learning
STEP-DOWN INHIBITORY
AVOIDANCE TASK
allocentrically using the extra maze cues. One possible and thereby render it insensitive to AP5 blockade in
A form of conditioning in which explanation for the lack of NMDAR-dependency in the hippocampus. A similar idea has been suggested
a rat is placed on a platform and pretrained animals is that the strategy or rule learning, by Moser and Moser57. The idea could be tested by
receives a shock when it steps off but not the spatial learning per se, requires NMDAR varying the timing of AP5 injection during training,
the platform. Memory for the
shock is measured as an
function. However, subsequent work did not support or by generating and analysing mice in which
increased latency to step off the this hypothesis49. So why does the MWM task depend on hippocampus-targeted inhibition of NR1 function is
platform on subsequent trials. hippocampal NMDARs? temporally controlled.

NATURE REVIEWS | NEUROSCIENCE VOLUME 5 | MAY 2004 | 3 6 5

©2004 Nature Publishing Group

REVIEWS
Day 1 Day 2
VI VI
CA1 V CA1 V
CA3 IV CA3 IV
III III
DG DG
II II
EC EC
Figure 3 | Impaired rapid formation of CA1 place fields as a consequence of CA3 NMDAR
knockout. On the first day of exploration in a novel environment, the spatial tuning of CA1 place
cells is compromised (that is, the place fields are enlarged) in CA3-NR1-knockout mice. It is
proposed that these CA1 responses reflect equally poorly tuned entorhinal cortex (EC) place-cell
activities. In normal animals, the poor spatial tuning will be quickly remedied within a few minutes
by the highly tuned drive from CA3 (not shown). But in the mutants, owing to the lack of NMDARs
(N-methyl-D-aspartate receptors), CA3 cells fail to provide such a drive and therefore the spatially
poorly tuned CA1 responses continue for at least 1 hour. However, by the time the mutants are
re-exposed to the same environment on the second day, a slow spatial refinement of CA1 place-
cell activities has been implemented off-line through an unknown mechanism. DG, dentate gyrus.
CA3 NMDARs and one-trial learning. The CA3 subfield
of the hippocampus has a robust recurrent network,
with pyramidal cells receiving synaptic contacts from
~2% of other CA3 pyramidal cells58,59 (FIG. 1). NMDAR-
dependent LTP has also been demonstrated at the
synapses of recurrent collaterals in CA3 as well as at
perforant path–CA3 synapses, whereas the plasticity
at dentate mossy fibre–CA3 synapses is NMDAR-
independent60–64. Marr and others have suggested that
recurrent networks with modifiable synaptic strength
could support the rapid acquisition of memories of a
one-time experience28,65,66. Although Steele and Morris’s
pharmacological blockade experiment indicated
that hippocampal NMDARs are involved in delay-
dependent rapid acquisition of one-time experience
memory, a more targeted intervention method would
be needed to identify specific hippocampal subfields
and circuitries in which NMDARs play this crucial part.
Lee and Kesner67 attempted to inject AP5 bilaterally,
primarily into CA3, CA1 or the DG in rats that had been
trained on a radial eight-arm maze using a delayed
non-matching-to-place experiment, and tested whether
the rats chose the previously unvisited arm to obtain a
reward using spatial cues. Rats that had been treated
with AP5 directed towards CA3 were impaired at carry-
ing out this task in a novel environment, consistent with
the hypothesis that NMDARs in CA3 are more impor-
tant than those in CA1 or DG for the acquisition of
spatial ‘working memory’ (equivalent to episodic-like
PAIRED-ASSOCIATE TASK memory) in a novel environment.
A task that involves the arbitrary Nakazawa et al. generated a mouse strain (CA3-NR1-
association of two stimuli (such
knockout mice) in which the NR1 gene deletion was
as word pairs in humans, or a
place and an odour or food item restricted to the CA3 pyramidal cells of adult mice68.
in animals). After exposure to These mutant mice were impaired in the DMP version
the pair, the subject is presented of the MWM task that was conducted with a 5-min
with one stimulus and tested for intertrial interval. By contrast, mutant mice performed
recall of the second. It can be
used to test declarative memory
normally when previously experienced platform loca-
in humans, or ‘episodic-like’ tions were used and in the standard version of the
memory in animals. MWM. These results showed that CA3 NMDARs are
366 | MAY 2004 | VOLUME 5
©2004 Nature Publishing Group
crucial for the rapid acquisition of one-trial memory
of novel spatial information69. Spatial tuning of CA1
place cells in mutant mice was normal in a familiar
environment but was significantly impaired (with
enlarged place fields and an augmented integrated
firing rate) in a novel environment69. Strikingly,
when the mutants were re-exposed to the ‘novel’
environment after being held in a home cage for 1 day,
spatial tuning was normal. So, the tuning deficit is seen
only during a defined period (at least 1 hour) after
exposure to a novel environment, and some consolida-
tion process occurs in the absence of the continuous
environmental information, resulting in normal
tuning by the next day.
It has been suggested that entorhinal neurons
provide an important input to CA1, particularly during
tasks that require the encoding of novel informa-
tion70–72. FIGURE 3 shows a model for the role of
NMDARs in CA3 in the rapid establishment of highly
tuned CA1 place cells when animals are exposed to
novel spatial information. During exposure to a novel
context, the CA1 response is initially and transiently
driven by direct input from the EC, which is spatially
broadly tuned73–75. In control animals, NMDAR activity
in CA3, perhaps operating through Hebbian recurrent
connections, allows the rapid formation of CA1 place
cells as the input from CA3 comes to dominate and
shape the EC input. This shift of dominant input
from EC to CA3 occurs so rapidly that it is difficult
to observe the initial, less tuned state of CA1 place
cells. By contrast, in CA3-NR1-knockout mice, as
a consequence of CA3 NMDAR ablation, the less
spatially-tuned EC place cells continue to drive CA1
place-cell activity for at least one hour after exposure to
a novel space. Subsequently, however, by the time the
mutants are re-exposed to the same context 1 day later,
a slow spatial refinement of CA1 place fields seems to
have been implemented off-line. It is not known which
hippocampal circuitries and synapses are involved in
this slow refinement process. However, knife cuts of
the connections between CA3 and CA1 do not block
reactivation of CA1 place cells on re-exposure to
a familiar environment72. This suggests that the
temporoammonic pathway and its synaptic plasticity
might be sufficient for this process76. Although CA3
NMDARs seem to be crucial for the rapid formation of
highly tuned CA1 place cells, their ablation does not
seem to block encoding of the novel information
entirely. If it did, it would be difficult to explain the
presence of highly tuned place cells on re-exposure on
day 2 (REF. 69).
Overall, the combined behavioural and physiological
analyses of the CA3-NR1-knockout mice and the
behavioural analysis of AP5-treated rats indicate that
CA3 NMDARs are crucial for rapid hippocampal
encoding of novel information, which is necessary for
rapid learning of one-time experience. A recent study on
a new PAIRED-ASSOCIATE TASK is consistent with this conclu-
sion, reporting that the encoding of a ‘what–where’
association between a specific odour and a location
depends on hippocampal NMDARs77.
www.nature.com/reviews/neuro

Memory consolidation
It is commonly thought that input from the external
world is initially encoded as nascent neural firing patterns
and is later stored in a more persistent form. Ebbinghaus78
first mentioned this phenomenon as the ‘overlearning
effect’, and Müller and Pilzecker79 proposed this as a
‘consolidation theory’, an idea later refined by Hebb10.
Memory is labile for a short time after its acquisition, and
treatments such as protein synthesis inhibition, electro-
convulsive shock or hypothermia can lead to memory
loss80–82. Two main mechanisms have been proposed for
‘memory consolidation’. One,‘trace-transfer consolida-
tion’, emphasizes the transfer of a memory trace over
time, for instance, from the hippocampus to the cortex1.
This between-structure transfer is sometimes referred to
as ‘systems consolidation’65,66,83–86. The second — cellular
consolidation — focuses more on the molecular events in
cells in a given brain region. It has been suggested that the
initially fragile memory trace is made more permanent
through biochemical and morphological synaptic and
cellular changes that mark the transition from short-term
to long-term forms of plasticity87–90. System consolidation
and cellular consolidation are not mutually exclusive.
Rather, it is likely that cellular consolidation occurs
in both pre-transfer (for example, hippocampus) and
post-transfer (for example, neocortex) brain systems82.
Briefly, cellular consolidation is initiated by the acti-
vation of NMDA, AMPA and metabotropic glutamate
receptors, and involves changes in the levels of second
messengers, followed by enhanced activity of protein
kinases such as PKA, PKC, PKG and the calmodulin
(CaM) kinases, as well as protein phosphatases such as
PP1, PP2A and PP2B. These molecular events lead to
de novo transcription through the activation of constitu-
tive and inducible transcriptional factors such as the
cyclic-AMP-responsive element-binding (CREB) family
of proteins91–100. Additional regulation can also occur at
the level of mRNA transport and translation101–103.
Post-training infusion of NMDAR antagonist. To exam-
ine the role of NMDARs in memory consolidation,
Morris et al. infused AP5 into the ventricles of rats that
had been trained on the hidden-platform MWM task. A
dose that impaired learning when administered before
training had no effect when administered 1 day after
training32,36. These data support the notion that
NMDARs are needed for memory acquisition but not for
consolidation. Further support for this idea came from
experiments in which AP5 and MK801 were adminis-
tered intra-ventricularly and systemically, respectively,
4 days after the last training day without effect104, and
from a study in which post-training administration of
MK801 had no effect on a probe trial 24 h later105. On
the other hand, Packard and Teather106,107 reported that
injection of AP5 or MK801 into rat hippocampus imme-
diately after training on the MWM task impaired the rats’
memory when they were tested 24 h later. These authors
defined the NMDAR-dependent time window to be less
than 2 h after completion of the training; however they
used an unusual training protocol that consisted of
a single session composed of eight continuous trials.
NATURE REVIEWS | NEUROSCIENCE
©2004 Nature Publishing Group
REVIEWS
In all of these studies, both the demands of the task and
the drug administration methods varied greatly and it is
not easy to compare the results and produce a coherent
interpretation.
In studies that seemed to show a function for
NMDARs in memory consolidation, no attempt was
made to restrict NMDAR blockade to a particular brain
system. Therefore, it is impossible to decide whether the
effects were exerted in the neocortex or hippocampus.
Shimizu et al.108 attempted to address this issue by com-
bining the tTA (tetracycline-dependent transactivator
protein)109 and Cre/loxP techniques to develop a mouse
line in which NMDAR ablation was both regionally and
temporally controlled. Treatment with doxycycline 1–7
days after training caused inhibition of NR1 expression in
hippocampal CA1 pyramidal cells and an impairment in
memory. Delaying NR1 inhibition to 9–14 days after
training prevented the memory impairment. The authors
concluded that the reactivation of NMDARs in the
hippocampus (that is, in CA1) is necessary for cellular
consolidation108. However, the restriction of NMDAR
inhibition to CA1 in this mouse strain was not clearly
demonstrated in their study.A recent analysis of the CA1-
NR1-knockout mice that were used in this study showed
that the absence of the NR1 subunit spread to the
neocortex after 2 months of age42, so it is premature to
conclude that reactivation of hippocampal NMDARs is
needed during memory consolidation.
Some studies have hinted at the involvement of
NMDARs in cellular memory consolidation, but further
investigation is needed to establish the role of hippo-
campal NMDAR function in consolidation. It should
also be noted that the hidden-platform MWM is not an
ideal paradigm for dissociating acquisition and consoli-
dation, because multiple training trials lasting several
days will inevitably make the two phases overlap. A para-
digm such as contextual fear conditioning, in which one
to a few training trials are sufficient for acquisition,
would be a method of choice.
Memory retrieval
In the aforementioned studies on the role of NMDARs
in cellular consolidation, the effect of an intervention
treatment on consolidation as opposed to retrieval was
distinguished by the duration between the time of the
completion of the training session and the time of
the treatment. This requires us to assume that a memory
test conducted soon after the completion of training
measures consolidation and/or retrieval, but a test con-
ducted long after training tests only retrieval. Based on
this assumption, there is accumulating evidence that the
ablation of hippocampal NMDAR function has little
effect on memory retrieval. Early evidence for this came
from a hippocampus-dependent olfactory discrimina-
tion task110 and a spatial reference memory task32,36. Later,
as mentioned above, similar results were obtained by
pharmacological106 and genetic intervention108. Other
behavioural tasks have also shown a role for hippocampal
NMDARs in encoding but not in retrieval. These include
inhibitory avoidance111–114 with the memory tests con-
ducted immediately after or 24 h after the end of training.
VOLUME 5 | MAY 2004 | 3 6 7
REVIEWS

a Full cue DG/EC input b Full cue DG/EC input In the hidden-platform MWM task, the training
EC input EC input (memory acquisition) and probe (memory retrieval)
trials are usually carried out in the same environment16,40.
SC SC Although this protocol provided basic insights into the
mechanisms of memory formation, it is not optimal for
CA1 CA3 CA1 CA3 investigating the mechanism that underlies memory
recall by pattern completion. Nakazawa et al. trained
Firing rate

Firing rate
CA3-NR1-knockout mice in the hidden-platform MWM
Position RC Position RC task under full-cue conditions and then subjected them
Control Mutant to probe trials either under the same full-cue conditions
c Partial cue DG/EC input d Partial cue DG/EC input or under conditions in which some of the cues were
EC input EC input removed (degraded-cue conditions)68. Although the
mutant mice acquired the spatial reference memory and
SC SC retrieved it normally under the full-cue conditions,
they were impaired at retrieving the memory under
CA1 CA3 CA1 CA3 degraded-cue conditions, unlike control mice. These data
indicated that CA3 NMDARs, in contrast to CA1
Firing rate

Firing rate

NMDARs41, were not necessary for the acquisition of

Position RC Position RC
Control Mutant
Figure 4 | Spatial pattern completion model in hippocampal networks. In control animals,
during memory acquisition under full-cue conditions, both recurrent collateral (RC) synapses within
CA3 and Schaffer collateral (SC) synapses in CA1 are modified in an NMDAR (N-methyl-D-aspartate
receptor)-dependent manner, leading to storage of the memory trace (small red dots) in CA3 and
CA1, respectively (a and c). In CA3-NR1-knockout mice this will happen in CA1 but not in CA3
(b and d). During the recall phase, if the full set of cues is presented, CA1 memory traces will be fully
activated in both control and mutant animals, leading to full recall (a and b). If only a partial set of
cues is provided to control animals, it will initially activate only partial CA3 memory traces, but
because of the robust connections among CA3 cells, activation of the entire CA3 memory traces will
be accomplished subsequently by iteration of activity within the recurrent CA3 network. This will, in
turn, activate the entire CA1 memory trace (c). In mutants, because of the lack of memory traces in
the recurrent CA3 network, which prevents iteration-mediated trace activation, CA1 memory traces
will not be fully activated (d). So, mutants exhibit impaired recall under partial-cue conditions. Vertical
arrows designate inputs from recall cues. Red dots with circles indicate reactivated memory trace
components68. DG, dentate gyrus; EC, entorhinal cortex. Modified, with permission, from REF. 68 
(2002) American Association for the Advancement of Science.
A likely reason for the dispensability of NMDARs
in retrieval comes from evidence that blockade of
hippocampal CA1 NMDARs has little effect on
AMPAR-mediated fast synaptic transmission11. So, even
if NMDAR blockade could completely freeze the
existing pattern of synaptic weights in the hippo-
campus, the cells would be able to fire and to faithfully
transmit the established network pattern using
AMPA/kainate receptors115,116.
CA3 NMDARs for associative memory recall. Marr
suggested that the recall process could occur either in one
step — ‘simple recall’ — or in a series of steps involving
recurrent network activation — a ‘collateral effect’28.
Personal experiences in daily life are unique and rarely
repeat exactly28,117. So, input patterns that change from
moment to moment would be able to reactivate only part
of a stored memory and would be unable to activate the
whole pattern without repeated iterations of firing in the
recurrent network. This concept is called ‘pattern comple-
tion’28. The presence of massive recurrent collaterals
with highly modifiable synapses made area CA3 an
attractive candidate site for the biological implementation
of this process, which is required for the retrieval of
hippocampus-dependent associative memory 26,118–122.
spatial reference memory that could be retrieved under
full-cue conditions, but that CA3 NMDARs were impor-
tant for recalling the spatial reference memory when only
a fraction of the cues were available. Recordings from
place cells in area CA1, downstream of CA3, corroborated
these behavioural data; when CA3-NR1-knockout mice
were allowed to explore a full-cue environment repeat-
edly, place cells with normal spatial tuning properties
were formed and they were fully reactivated when the
animals were returned to the environment after several
hours in a home cage. By contrast, when the mutants
were returned to an environment in which only
a fraction of the original cues were available, the place-
cell reactivation was severely impaired. These behavioural
and physiological data, along with the finding that
CA1-NR1-knockout mice showed a severe deficit in the
acquisition of the hidden-platform MWM task40
and uncoordinated, spatially less-tuned CA1 place-cell
activity41, led Nakazawa et al. to conclude that CA1
NMDARs are essential for the acquisition of spatial
reference memory but that CA3 NMDARs are not.
Instead, CA3 NMDARs are crucial for the retrieval of the
memory under conditions in which pattern completion
is required.According to this proposal, the primary deficit
in CA3-NR1-knockout mice was in the formation
of NMDAR-dependent ensemble memory traces in
CA3. This manifested as a deficit in recall by pattern
completion (FIG. 4).
NMDAR function and place cells
Since O’Keefe and Nadel123 first proposed that the hippo-
campus is a neural substrate of a ‘cognitive map’ and
described the mnemonic aspects of place-cell activity,
tremendous effort has been made to determine whether
place-cell activity is directly related to the formation and
expression of spatial memory traces. BOX 3 lists the condi-
tions that must be met by any physiological process to be
called a memory trace. Many studies have provided
evidence that the pattern of place-cell activities can
be viewed as an expression of a memory trace at the
neuronal ensemble level123–130. Most convincingly, place
cells show spatially characteristic stable firing patterns
across temporally separated recording sessions — a

368 | MAY 2004 | VOLUME 5 www.nature.com/reviews/neuro

©2004 Nature Publishing Group


Box 3 | Necessary attributes of memory traces
Experience dependency (criterion 1)
The memory trace should form in an experience-dependent manner.
Information specificity (criterion 2)
The trace should be specific to the information that is acquired.
Persistency (criterion 3)
The trace must outlast the period during which the animal is exposed to the information.
Ability to be reactivated (criterion 4)
The subsequent presentation of at least part of the original cues should reactivate the trace.
unique set of place cells is reactivated on subsequent
exposure to the same environment, and these patterns
can be subsequently reactivated independently of behav-
iour124,127,128,131–133. In addition, the entire set of place cells
can be reactivated by a subset of the original set of
cues68,134–137. These features of place cells satisfy criteria 2, 3
and 4 in BOX 3. However, it has been difficult to show
directly that place-cell formation is experience dependent
(criterion 1). It has been shown that in a novel space the
ensemble code, which is initially less robust, improves
rapidly with exploration; however, this change is so rapid
that improvements in the coding of individual cells have
been impossible to monitor125. Although Mehta et al.
have reported lap-dependent modifications of preexisting
place fields138,139, the initial changes in place-field size that
might represent the acquisition of novel spatial informa-
tion have not been fully characterized. The ablation of
NR1 in area CA3 slows down the de novo formation of
CA1 place fields69.Although the kinetics and mechanisms
of the maturation of these place fields have not been fully
characterized, the enlarged place fields that were observed
during the first hour of exposure became condensed to
normal sizes in the home cage before reexposure to the
same environment the next day. If these enlarged place
fields represent an early, transient state of place-cell for-
mation, this finding supports the experience dependency
of place-cell formation, and therefore the role of these
cells in a memory trace.
Other groups have used pharmacology to address the
role of the NMDAR in the hippocampal encoding of
space. Kentros et al. examined the differential effects
of systemic administration of the NMDAR antagonist
CPP (3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic
acid) on the formation and maintenance of CA1 place
cells140. The global blockade of NMDARs had no effect
on the formation and short-term stability of CA1 place
cells in a novel environment, nor on the reactivation of
place cells in a familiar environment. The observation
of normal place fields in novel space in the absence
of NMDAR function seems to contradict the
finding of enlarged and diffuse CA1 place fields in
CA3-NR1-knockout mice69.
Ekstrom et al.141 used pharmacological blockade to
test whether NMDAR-mediated plasticity is necessary for
the experience-dependent asymmetric shift of established
CA1 place fields139. Treatment with CPP caused a deficit
in the lap-dependent skewing of fields and, as a result,
slightly smaller fields. This reduction of place-field sizes
also contradicts the enlarged and diffuse CA1 fields of
NATURE REVIEWS | NEUROSCIENCE
©2004 Nature Publishing Group
REVIEWS
CA1-NR1-knockout mice in familiar environments41.
One possible explanation for the differences in the spatial
tuning properties of place cells in pharmacologically
treated and genetically engineered rodents is that the
genetic interventions (even if they occur only postnatally)
are not as acute as the pharmacological ones. So, the NR1
deletion might have led to an accumulation of cellular
and/or system level changes, and these might have con-
tributed to the phenotype. Conversely, the genetic lesion
was more targeted and more complete, so it might have
revealed more specific deficits. For instance, the establish-
ment of place fields could require a balance of excitation
and inhibition on both the cellular and circuit levels.
Systemic administration of CPP is expected to inhibit
NMDARs on both excitatory and inhibitory cells
involved in place-field formation, possibly in a coordi-
nated fashion, leading to the apparently normal pheno-
type. By contrast, the genetic NR1 blockade was restricted
to the excitatory neurons, and so might have led to the
impairment. The future application of cell-type-specific,
rapidly inducible genetic manipulation could address
these issues and clarify the matter.
Although the study of Kentros et al.140 lacks regional
specificity, it does have the advantage of precise temporal
control and therefore has shed light on the relationship
between NMDARs and the stability of CA1 place fields.
Place fields that formed de novo in the presence of CPP
were unstable, but CPP had no effect on the maintenance
of previously established place fields140. These results are
consistent with behavioural evidence that hippocampal
NMDARs are required for the acquisition and perhaps
the initial consolidation of hippocampus-dependent
memory, but not for its maintenance. Other studies with
transgenic mice in which signalling and transcriptional
molecules downstream of the NMDAR were altered also
showed similar effects on the stability of place fields142–145.
However, one should apply caution in interpreting these
data because it is technically difficult to maintain stability
in the electrophysiological isolation and identification of
cells recorded across days in these long-term recording
experiments.
Conclusions
We have summarized recent advances in our under-
standing of the roles of NMDARs in the acquisition,
consolidation and recall of hippocampus-dependent
memory with a focus on spatial memory. Behavioural
studies of rodents in which NMDAR function is
blocked by pharmacological or genetic manipulations
indicate that hippocampal NMDARs are crucial for
the acquisition of hippocampus-dependent memory,
and particularly for ‘episodic-like’ memory. Whether
hippocampal NMDARs are involved in the cellular con-
solidation of memory remains to be determined, but
they do not seem to be directly involved in memory
retrieval. Cell-type-restricted gene ablation techniques
are beginning to reveal distinct mnemonic functions of
NMDARs in the different hippocampal areas and
circuits. So, although NMDARs in CA1 pyramidal cells
seem to be crucial for the acquisition of spatial reference
memory, NMDARs in CA3 pyramidal cells are not.
VOLUME 5 | MAY 2004 | 3 6 9
REVIEWS

Instead, CA3 NMDARs seem to be important for the
rapid acquisition of one-trial memory (episodic-like
memory) and for the retrieval of associative memory
by pattern completion.
Although these behavioural studies with pharma-
cologically or genetically engineered rodents have
contributed greatly to our understanding of the impor-
tance of NMDAR-dependent synaptic plasticity in
hippocampus-dependent memory, it is unlikely that
a memory trace resides at an individual synapse with
modified transmission efficacy. As Hebb proposed,
memory is likely to be encoded in the pattern of activity
of a cellular ensemble, the individual components of
which are connected by synapses with modified synaptic
efficacy10. In several studies in which the NR1 gene
knockout was targeted to hippocampal CA1 or CA3
pyramidal cells of adult mice, the spatial tuning qualities
of CA1 place cells were determined under conditions
that mimicked those under which behavioural aspects of
hippocampal-dependent learning and memory were
examined. The deficits in the spatial tuning quality of
place cells, both at the individual and ensemble cell
levels, correlated well with the impairments of behav-
iourally assessed learning and memory. In addition, the
analyses of the abilities of CA3-NR1-knockout mice in
rapid one-trial learning and in the encoding of a novel
space led to evidence for experience-dependency of
place-cell formation, an important attribute of an
ensemble memory trace. The effects of NMDAR antago-
nists on the long-term stability of CA1 place cells
correlated well with the effects of these agents on the
acquisition and consolidation of spatial memory. All of
these studies support the idea that hippocampal place
cells are an excellent candidate for an ensemble memory
trace. However, most of the evidence that led to this
notion concerns spatial encoding and spatial memory
and in almost all of these studies, the recording of place
cells have been carried out, for technical reasons, in
setups that mimic, but are not identical to, those of the
learning and memory tasks. So, one challenge for future
studies is to devise a paradigm with which memory and
place-cell activities can be assessed simultaneously. There
is also still much to be learned about how information
relating to non-spatial parameters, such as behavioural
state, motivation and olfactory and other sensory inputs,
is encoded by and represented in place-cell activity146–151.
With further multidisciplinary studies consisting of
conditional transgenic technology, in vivo multi-unit
recording and clever behavioural paradigms, we might
move closer to our ultimate goal of understanding the
molecular, cellular and neuronal ensemble mechanisms
for hippocampus-dependent learning and memory.

1. Scoville, W. B. & Milner, B. Loss of recent memory after
bilateral hippocampal lesions. J. Neurol. Neurosurg.
Psychiatry 20, 11–21 (1957).
2. Schmolck, H., Kensinger, E. A., Corkin, S. & Squire, L. R.
Semantic knowledge in patient H. M. and other patients with
bilateral medial and lateral temporal lobe lesions.
Hippocampus 12, 520–533 (2002).
3. Rempel-Clower, N. L., Zola, S. M., Squire, L. R. & Amaral, D. G.
Three cases of enduring memory impairment after bilateral
damage limited to the hippocampal formation. J. Neurosci.
16, 5233–5255 (1996).
4. Zola, S. M. & Squire, L. R. Relationship between magnitude
of damage to the hippocampus and impaired recognition
memory in monkeys. Hippocampus 11, 92–98 (2001).
5. Jarrard, L. E. On the role of the hippocampus in learning and
memory in the rat. Behav. Neural Biol. 60, 9–26 (1993).
6. Morris, R. G. M., Garrud, P., Rawlins, J. N. P. & O’Keefe,
J. Place navigation impaired in rats with hippocampal
lesions. Nature 297, 681–683 (1982).
7. Silva, A. J., Giese, K. P., Fedorov, N. B., Frankland, P. W. &
Kogan, J. H. Molecular, cellular, and neuroanatomical
substrates of place learning. Neurobiol. Learn. Mem. 70,
44–61 (1998).
8. O’Keefe, J. & Dostrovsky, J. The hippocampus as a spatial
map. Preliminary evidence from unit activity in the freely-
moving rat. Brain Res. 34, 171–175 (1971).
9. Bliss, T. V. P. & Lømo, T. Long-lasting potentiation of
synaptic transmission in the dentate area of the
anaesthetized rabbit following stimulation of the perforant
path. J. Physiol. (Lond.) 232, 331–356 (1973).
10. Hebb, D. O. The Organization of Behavior: A
Neuropsychological Theory (John Wiley, New York, 1949).
11. Collingridge, G. L., Kehl, S. J. & McLennon, H. Excitatory
amino acids in synaptic transmission in the Schaffer
collateral-commissural pathway of the rat hippocampus.
J. Physiol. (Lond.) 334, 33–46 (1983).
The first demonstration that long-term potentiation in
the hippocampus is NMDAR-dependent.
12. Mayer, M. L., Westbrook, G. L. & Guthrie, P. B. Voltage-
dependent block by Mg2+ of NMDA responses in spinal cord
neurones. Nature 309, 261–263 (1984).
13. MacDermott, A. B., Mayer, M. L., Westbrook, G. L., Smith,
S. J. & Barker, J. L. NMDA-receptor activation increases
cytoplasmic calcium concentration in cultured spinal cord
neurones. Nature 321, 519–522 (1986).
14. Nowak, L., Bregestovski, P., Ascher, P., Herbert, A. &
Prochiantz, A. Magnesium gates glutamate-activated channels
in mouse central neurones. Nature 307, 462–465 (1984).
This study utilized the patch-clamp technique to
provide the first evidence of a voltage-dependent
Mg2+ block of the NMDA Ca2+ channel
15. Lester, R. A., Clements, J. D., Westbrook, G. L. & Jahr, C. E.
Channel kinetics determine the time course of NMDA
receptor-mediated synaptic currents. Nature 346, 565–567
(1990).
16. Morris, R. G. M., Anderson, E., Lynch, G. S. & Baudry, M.
Selective impairment of learning and blockade of long-term
potentiation by an N-methyl-D-aspartate receptor
antagonist, AP5. Nature 319, 774–776 (1986).
This paper was the first to show that NMDAR activity
is needed for spatial learning and to draw parallels
between NMDAR function, hippocampal synaptic
plasticity and spatial learning.
17. Moriyoshi, K. et al. Molecular cloning and characterization of
the rat NMDA receptor. Nature 354, 31–37 (1991).
The first identification of a gene encoding a subunit of
the NMDAR, which paved the way for much of the
later genetic and structural work.
18. Dunah, A. W. et al. Biochemical studies of the structure and
function of the N-methyl-D-aspartate subtype of glutamate
receptors. Mol. Neurobiol. 19, 151–179 (1999).
19. Cull-Candy, S., Brickley, S. & Farrant, M. NMDA receptor
subunits: diversity, development and disease. Curr. Opin.
Neurobiol. 11, 327–335 (2001).
20. Li, Y., Erzurumulu, R., Chen, C., Jhaveri, S. & Tonegawa, S.
Whisker-related neuronal patterns fail to develop in the
trigeminal brainstem nuclei of NMDAR1 knockout mice. Cell
76, 427–437 (1994).
21. Forrest, D. et al. Targeted disruption of NMDA receptor 1
gene abolishes NMDA response and results in neonatal
death. Neuron 13, 325–338 (1994).
22. Amaral, D. G. & Witter, M. P. The three-dimensional
organization of the hippocampal formation: a review of
anatomical data. Neuroscience 31, 571–591 (1989).
23. Miles, R. & Wong, R. K. Excitatory synaptic interactions
between CA3 neurones in the guinea-pig hippocampus.
J. Physiol. (Lond.) 373, 397–418 (1986).
24. MacVicar, B. A. & Dudek, F. E. Local synaptic circuits in rat
hippocampus: interactions between pyramidal cells. Brain
Res. 184, 220–223 (1980).
25. McClelland, J. L. & Goddard, N. H. Considerations arising
from a complementary learning systems perspective on
hippocampus and neocortex. Hippocampus 6, 654–665
(1996).
26. McNaughton, B. L. & Morris, R. G. M. Hippocampal
synaptic enhancement and information storage within a
distributed memory system. Trends Neurosci. 10, 408–415
(1987).
27. O’Reilly, R. C. & McClelland, J. L. Hippocampal conjunctive
encoding, storage, and recall: avoiding a trade-off.
Hippocampus 4, 661–682 (1994).
28. Marr, D. Simple memory: a theory for archicortex. Philos.
Trans. R. Soc. London B 262, 23–81 (1971).
A seminal work that laid the foundation of much of our
current understanding of hippocampal computation.
Marr’s models include a discussion of partial-cue-
based recall and experience-dependent modifications
of synaptic strength.
29. O’Reilly, R. C. & Rudy, J. W. Conjunctive representations in
learning and memory: principles of cortical and hippocampal
function. Psychol. Rev. 108, 311–345 (2001).
30. Morris, R. G. M. Spatial localization does not require the
presence of local cues. Learn. Motiv. 12, 239–260 (1981).
31. Davis, S., Butcher, S. P. & Morris, R. G. M. The NMDA
receptor antagonist D-2-amino-5-phosphonopentanoate
(D-AP5) impairs spatial learning and LTP in vivo at
intracerebral concentrations comparable to those that block
LTP in vitro. J. Neurosci. 12, 21–34 (1992).
32. Morris, R. G. M. Synaptic plasticity and learning: selective
impairment of learning in rats and blockade of long-term
potentiation in vivo by the N-methyl-D-aspartate receptor
antagonist AP5. J. Neurosci. 9, 3040–3057 (1989).
33. Riedel, G. et al. Reversible neural inactivation reveals
hippocampal participation in several memory processes.
Nature Neurosci. 2, 898–905 (1999).
34. Abraham, W. C. & Kairiss, E. W. Effects of the NMDA
antagonist 2AP5 on complex spike discharge by hippocampal
pyramidal cells. Neurosci. Lett. 89, 36–42 (1988).
35. Errington, M. L., Lynch, M. A. & Bliss, T. V. Long-term
potentiation in the dentate gyrus: induction and increased
glutamate release are blocked by
D(–)aminophosphonovalerate. Neuroscience 20, 279–284
(1987).
36. Morris, R. G. M., Davis, S. & Butcher, S. P. Hippocampal
synaptic plasticity and NMDA receptors: a role in information
storage? Phil. Trans. R. Soc. Lond. B 329, 187–204 (1990).
37. Hargreaves, E. L. & Cain, D. P. Hyperactivity, hyper-reactivity,
and sensorimotor deficits induced by low doses of the
N-methyl-D-aspartate non-competitive channel blocker
MK801. Behav. Brain Res. 47, 23–33 (1992).
38. Sternberg, N., Hamilton, D. & Hoess, R. Bacteriophage P1
site-specific recombination. II. Recombination between loxP
and the bacterial chromosome. J. Mol. Biol. 150, 487–507
(1981).

370 | MAY 2004 | VOLUME 5 www.nature.com/reviews/neuro

©2004 Nature Publishing Group


39. Hoess, R. H., Ziese, M. & Sternberg, N. P1 site-specific
recombination: nucleotide sequence of the recombining
sites. Proc. Natl Acad. Sci. USA 79, 3398–3402 (1982).
40. Tsien, J. Z., Huerta, P. T. & Tonegawa, S. The essential role
of hippocampal CA1 NMDA receptor-dependent synaptic
plasticity in spatial memory. Cell 87, 1327–1338 (1996).
This paper offered the first genetic demonstration of
the effect of a brain-subregion-restricted NMDA
receptor knockout on spatial memory.
41. McHugh, T. J., Blum, K. I., Tsien, J. Z., Tonegawa, S. & Wilson,
M. A. Impaired hippocampal representation of space in CA1-
specific NMDAR1 knockout mice. Cell 87, 1339–1349 (1996).
This study was the first to show impaired coherent
firing of CA1 place cells using genetically engineered
NMDAR-knockout mice, demonstrating the powerful
combination of conditional transgenic technology and
multi-electrode recording.
42. Fukaya, M., Kato, A., Lovett, C., Tonegawa, S. & Watanabe, M.
Retention of NMDA receptor NR2 subunits in the lumen of
endoplasmic reticulum in targeted NR1 knockout mice.
Proc. Natl Acad. Sci. USA 100, 4855–4860 (2003).
43. Huerta, P. T., Sun, L. D., Wilson, M. A. & Tonegawa, S.
Formation of temporal memory requires NMDA receptors
within CA1 pyramidal neurons. Neuron 25, 473–480 (2000).
44. Rondi-Reig, L., Libbey, M., Eichenbaum, H. & Tonegawa, S.
CA1-specific N-methyl-D-aspartate receptor knockout mice
are deficient in solving a nonspatial transverse patterning
task. Proc. Natl Acad. Sci. USA 98, 3543–3548 (2001).
45. Rampon, C. et al. Enrichment induces structural changes and
recovery from nonspatial-memory deficits in CA1-NMDAR1-
knockout mice. Nature Neurosci. 3, 238–244 (2000).
46. Bannerman, D. M., Good, M. A., Butcher, S. P., Ramsay, M.
& Morris, R. G. Distinct components of spatial learning
revealed by prior training and NMDA receptor blockade.
Nature 378, 182–186 (1995).
47. Saucier, D. & Cain, D. P. Spatial learning without NMDA
receptor-dependent long-term potentiation. Nature 378,
186–189 (1995).
References 46 and 47 provided strong evidence that
NMDARs could be made dispensable for learning the
water maze task if animals were pretrained on an
identical spatial or even similar, non-spatial task. This
was the first real evidence that the water maze task
might involve several types of learning, some of which
require NMDARs in the hippocampus and some of
which do not.
48. Roesler, R. et al. Intrahippocampal infusion of the NMDA
receptor antagonist AP5 impairs retention of an inhibitory
avoidance task: protection from impairment by pretraining or
preexposure to the task apparatus. Neurobiol. Learn. Mem.
69, 87–91 (1998).
49. Hoh, T., Beiko, J., Boon, F., Weiss, S. & Cain, D. P. Complex
behavioral strategy and reversal learning in the watermaze
without NMDA receptor-dependent long-term potentiation.
J. Neurosci. 19, RC2(1–5) (1999).
50. Steele, R. J. & Morris, R. G. M. Delay-dependent impairment
of a matching-to-place task with chronic and
intrahippocampal infusion of the NMDA-antagonist D-AP5.
Hippocampus 9, 118–136 (1999).
This study showed that a delayed matching-to-place
task in the water maze requires hippocampal NMDA
receptors in rats, providing new insights into the
involvement of the hippocampus in episodic-like
memory.
51. Tulving, E. in Organization of Memory (eds Tulving, E. &
Donaldson, W.) 381–403 (Academic, New York, 1972).
52. Tulving, E. Episodic memory: from mind to brain. Annu. Rev.
Psychol. 53, 1–25 (2002).
53. Tulving, E. & Markowitsch, H. J. Episodic and declarative
memory: role of the hippocampus. Hippocampus 8,
198–204 (1998).
54. Clayton, N. S. & Dickinson, A. Episodic-like memory during
cache recovery by scrub jays. Nature 395, 272–274 (1998).
Taking advantage of the natural caching behaviour of
the scrub jay, these authors provide the strongest
evidence for the existence of ‘what, where and when
(episodic-like) memory’ in animals.
55. Clayton, N. S., Bussey, T. J. & Dickinson, A. Can animals
recall the past and plan for the future? Nature Rev. Neurosci.
4, 685–691 (2003).
56. Morris, R. G. M. Episodic-like memory in animals:
psychological criteria, neural mechanisms and the value of
episodic-like tasks to investigate animal models of
neurodegenerative diseases. Philos. Trans. R. Soc. Lond. B
356, 1453–1465 (2001).
57. Moser, M. B. & Moser, E. I. Pretraining and the function of
hippocampal long-term potentiation. Neuron 26, 559–561
(2000).
58. Miles, R. & Traub, R. Excitatory synaptic interactions
between CA3 neurons in the guinea-pig hippocampus.
J. Physiol. (Lond.) 373, 397–418 (1986).
NATURE REVIEWS | NEUROSCIENCE
59. MacVicar, B. & Dudek, F. Local synaptic circuits in rat
hippocampus: interaction between pyramidal cells. Brain
Res. 184, 220–223 (1980).
60. Harris, E. W. & Cotman, C. W. Long-term potentiation of
guinea pig mossy fiber response is not blocked by N-methyl-
D-aspartate antagonist. Neurosci. Lett. 70, 132–137 (1986).
61. Williams, S. & Johnston, D. Muscarinic depression of long-
term potentiation in CA3 hippocampal neurons. Science
242, 84–87 (1988).
62. Zalutsky, R. A. & Nicoll, R. A. Comparison of two forms of
long-term potentiation in single hippocampal neurons.
Science 248, 1619–1624 (1990).
63. Berger, T. W. & Yeckel, M. F. in Long-Term Potentiation: A
Debate of Current Issues (eds Baudry, M. & Davis, J. L.)
327–356 (MIT Press, Cambridge, Massachusetts, 1991).
64. Berzhanskaya, J., Urban, N. N. & Barrionuevo, G.
Electrophysiological and pharmacological characterization
of the direct perforant path input to hippocampal area CA3.
J. Neurophysiol. 79, 2111–2118 (1998).
65. McClelland, J. L., McNaughton, B. L., O’Reilly, R. C. & Nadel,
L. Complementary roles of hippocampus and neocortex in
learning and memory. Soc. Neurosci. Abstr. 18, 1216 (1992).
66. McClelland, J. L., McNaughton, B. L. & O’Reilly, R. C. Why
there are complementary learning systems in the
hippocampus and neocortex: insights from the success and
failures of connectionist models of learning and memory.
Psychol. Rev. 102, 419–457 (1995).
The authors of references 65 and 66 formulated the
theoretical basis of complementary learning systems
of hippocampus and neocortex, which allows the
brain to rapidly learn new on-line events without
disrupting old off-line information, and to integrate
these new events properly with previously stored
experiences.
67. Lee, I. & Kesner, R. P. Differential contribution of NMDA
receptors in hippocampal subregions to spatial working
memory. Nature Neurosci. 5, 162–168 (2002).
68. Nakazawa, K. et al. Requirement for hippocampal CA3
NMDA receptors in acquisition and recall of associative
memory. Science 297, 211–218 (2002).
This study combines regionally restricted genetic
modification with in vivo physiology and behaviour to
provide strong evidence that NMDARs in region CA3
are required for memory recall by pattern completion.
69. Nakazawa, K. et al. Hippocampal CA3 NMDA receptors are
crucial for memory acquisition of one-time experience.
Neuron 38, 305–315 (2003).
70. Vinogradova, O. S. in The Hippocampus (eds Isaacson, R. L.
& Pribram, K. H.) 3–69 (Plenum, New York and London,
1975).
71. McNaughton, B. L., Barnes, C. A., Meltzer, J. & Sutherland,
R. J. Hippocampal granule cells are necessary for normal
spatial learning but not for spatially-selective pyramidal cell
discharge. Exp. Brain Res. 76, 485–496 (1989).
72. Brun, V. H. et al. Place cells and place recognition
maintained by direct entorhinal-hippocampal circuitry.
Science 296, 2243–2246 (2002).
73. Barnes, C. A., McNaughton, B. L., Mizumori, S. J., Leonard,
B. W. & Lin, L.-H. Comparison of spatial and temporal
characteristics of neuronal activity in sequential stages of
hippocampal processing. Prog. Brain Res. 83, 287–300
(1990).
74. Frank, L. M., Brown, E. N. & Wilson, M. A. Trajectory
encoding in the hippocampus and entorhinal cortex. Neuron
27, 169–178 (2000).
75. Quirk, G. J., Muller, R. U., Kubie, J. L. & Ranck, J. B. Jr. The
positional firing properties of medial entorhinal neurons:
description and comparison with hippocampal place cells.
J. Neurosci. 12, 1945–1963 (1992).
76. Moser, E. I. & Moser, M. B. One-shot memory in
hippocampal CA3 networks. Neuron 38, 147–148 (2003).
77. Day, M., Langston, R. & Morris, R. G. Glutamate-receptor-
mediated encoding and retrieval of paired-associate
learning. Nature 424, 205–209 (2003).
78. Ebbinghaus, H. Memory: A Contribution to Experimental
Psychology (Dover, New York, 1885).
79. Müller, G. E. & Pilzecker, A. Experimentelle Beitrage zur
Lehre vom Gedachtnis. Z. Psychol. Erganzungsband 1,
1–300 (1900).
80. McGaugh, J. L. Time-dependent processes in memory
storage. Science 153, 1351–1358 (1966).
81. McGaugh, J. L. Memory: a century of consolidation.
Science 287, 248–251 (2000).
82. Nader, K. Memory traces unbound. Trends Neurosci. 26,
65–72 (2003).
83. Squire, L. R. Memory and the hippocampus: a synthesis
from findings with rats, monkeys, and humans. Psychol.
Rev. 99, 195–231 (1992).
84. Squire, L. R. & Alvarez, P. Retrograde amnesia and memory
consolidation: a neurobiological perspective. Curr. Opin.
Neurobiol. 5, 169–177 (1995).
©2004 Nature Publishing Group
REVIEWS
85. Nadel, L. & Moscovitch, M. Memory consolidation,
retrograde amnesia and the hippocampal complex. Curr.
Opin. Neurobiol. 7, 217–227 (1997).
86. Bontempi, B., Lauren-Demir, C., Destrade, C. & Jaffard, R.
Time-dependent reorganization of brain circuitry underlying
long-term memory storage. Nature 400, 671–675 (1999).
87. Davis, H. P. & Squire, L. R. Protein synthesis and memory: a
review. Psych. Bull. 96, 518–559 (1984).
88. DeZazzo, J. & Tully, T. Dissection of memory formation: from
behavioral pharmacology to molecular genetics. Trends
Neurosci. 18, 212–218 (1995).
89. Bailey, C. H., Bartsch, D. & Kandel, E. R. Toward a
molecular definition of long-term memory storage. Proc.
Natl Acad. Sci. USA 93, 13445–13452 (1996).
90. Nader, K., Schafem, G. E. & LeDoux, J. E. The labile nature
of consolidation theory. Nature Rev. Neurosci. 1, 216–219
(2000).
91. Abel, T. & Kandel, E. Positive and negative regulatory
mechanisms that mediate long-term memory storage. Brain
Res. Rev. 26, 360–378 (1998).
92. Adams, J. P. & Sweatt, J. D. Molecular psychology: roles for
the ERK MAP kinase cascade in memory. Annu. Rev.
Pharmacol. Toxicol. 42, 135–163 (2002).
93. Dudai, Y. Molecular bases of long-term memories: a
question of persistence. Curr. Opin. Neurobiol. 12, 212–216
(2002).
94. Bozon, B. et al. MAPK, CREB and zif268 are all required for
the consolidation of recognition memory. Philos. Trans. R.
Soc. Lond. B 358, 805–814 (2003).
95. Kang, H. et al. An important role of neural activity-dependent
CaMKIV signaling in the consolidation of long-term memory.
Cell 106, 771–783 (2001).
96. Kida, S. et al. CREB required for the stability of new and
reactivated fear memories. Nature Neurosci. 5, 348–355
(2002).
97. Schafe, G. E., Nader, K., Blair, H. T. & LeDoux, J. E. Memory
consolidation of Pavlovian fear conditioning: a cellular and
molecular perspective. Trends Neurosci. 24, 540–546
(2001).
98. Taubenfeld, S. M., Milekic, M. H., Monti, B. & Alberini, C. M.
The consolidation of new but not reactivated memory
requires hippocampal C/EBPb. Nature Neurosci. 4,
813–818 (2001).
99. Abel, T. & Lattal, K. M. Molecular mechanisms of memory
acquisition, consolidation and retrieval. Curr. Opin.
Neurobiol. 11, 180–187 (2001).
100. Izquierdo, I. & Medina, J. H. Memory formation: the
sequence of biochemical events in the hippocampus and its
connection to activity in other brain structures. Neurobiol.
Learn. Mem. 68, 285–316 (1997).
101. Steward, O. & Schuman, E. M. Protein synthesis at synaptic
sites on dendrites. Annu. Rev. Neurosci. 24, 299–325
(2001).
102. Kelleher, R. J., Govindarajan, A., Jung H.-Y., Kang, H. &
Tonegawa, S. Translational control by MAPK signaling in
long-term synaptic plasticity and memory. Cell 116,
467–479 (2004).
103. Wu, L. et al. CPEB-mediated cytoplasmic polyadenylation
and the regulation of experience-dependent translation of α-
CaMKII mRNA at synapses. Neuron 21, 1129–1139 (1998).
104. Heale, V. & Harley, C. MK-801 and AP5 impair acquisition,
but not retention, of the Morris milk maze. Pharmacol.
Biochem. Behav. 36, 145–149 (1990).
105. Norris, C. M. & Foster, T. C. MK-801 improves retention in
aged rats: implications for altered neural plasticity in age-
related memory deficits. Neurobiol. Learn. Mem. 71,
194–206 (1999).
106. Packard, M. G. & Teather, L. A. Double dissociation of
hippocampal and dorsal-striatal memory systems by
posttraining intracerebral injections of 2-amino-5-
phosphonopentanoic acid. Behav. Neurosci. 111, 543–551
(1997).
107. Packard, M. G. & Teather, L. A. Posttraining injections of
MK-801 produce a time-dependent impairment of memory
in two water maze tasks. Neurobiol. Learn. Mem. 68, 42–50
(1997).
108. Shimizu, E., Tang, Y. P., Rampon, C. & Tsien, J. Z. NMDA
receptor-dependent synaptic reinforcement as a crucial
process for memory consolidation. Science 290,
1170–1174 (2001).
109. Gossen, M. & Bujard, H. Anhydrotetracycline, a novel effector
for tetracycline controlled gene expression systems in
eukaryotic cells. Nucleic Acids Res. 21, 4411–4412 (1993).
110. Stäubli, U., Thibault, O., DiLorenzo, M. & Lynch, G.
Antagonism of NMDA receptors impairs acquisition but not
retetion of olfactory memory. Behav. Neurosci. 103, 54–60
(1989).
111. Parada-Turska, J. & Turski, W. A. Excitatory amino acid
antagonists and memory: effect of drugs acting at N-methyl-
D-aspartate receptors in learning and memory tasks.
Neuropharmacology 29, 1111–1116 (1990).
VOLUME 5 | MAY 2004 | 3 7 1
REVIEWS
112. Venable, N. & Kelly, P. H. Effects of NMDA receptor
antagonists on passive avoidance learning and retrieval in
rats and mice. Psychopharmacology (Berl.) 100, 215–221
(1990).
113. Szapiro, G. et al. Participation of hippocampal metabotropic
glutamate receptors, protein kinase A and mitogen-
activated protein kinases in memory retrieval. Neuroscience
99, 1–5 (2000).
114. Liang, K. C., Hon, W. & Davis, M. Pre- and posttraining
infustion of N-methyl-D-aspartate receptor antagonists into
the amygdala impair memory in an inhibitory avoidance task.
Behav. Neurosci. 108, 241–253 (1994).
115. Martin, S. J., Grimwood, P. D. & Morris, R. G. M. Synaptic
plasticity and memory: an evaluation of the hypothesis.
Annu. Rev. Neurosci. 23, 649–711 (2000).
116. Martin, S. J. & Morris, R. G. M. New life in an old idea: the
synaptic plasticity and memory hypothesis revisited.
Hippocampus 12, 609–636 (2002).
117. Chomsky, N. Syntactic Structures (Mouton, The Hague,
1957).
118. Gardner-Medwin, A. R. The recall of events through the
learning of associations between their parts. Proc. R. Soc.
Lond. B 194, 375–402 (1976).
119. Hopfield, J. J. Neural networks and physical systems with
emergent collective computational abilities. Proc. Natl Acad.
Sci. USA 79, 2554–2558 (1982).
120. Rolls, E. T. in The Computing Neuron (eds Durbin, R., Maill,
C. & Mitchison, G.) 125–159 (Addison-Wesley, Wokingham,
UK, 1989).
121. Willshaw, D. J. & Buckingham, J. T. An assessment of
Marr’s theory of the hippocampus as a temporary memory
storage. Phil. Trans. R. Soc. Lond. B 329, 205–215 (1990).
122. Hasselmo, M. E., Schnell, E. & Barkai, E. Dynamics of
learning and recall at excitatory recurrent synapses and
cholinergic modulation in rat hippocampal region CA3.
J. Neurosci. 15, 5249–5262 (1995).
123. O’Keefe, J. & Nadel, L. The Hippocampus as a Cognitive
Map (Clarendon, Oxford, 1978).
This book uses both the animal and the human
literature to outline a theory concerning the role of the
hippocampus in spatial representation. The ideas
outlined in this work have been the basis for
countless experiments in the field and continue to
generate important discussion.
124. Wilson, M. A. & McNaughton, B. L. Reactivation of
hippocampal ensemble memories during sleep. Science
265, 676–679 (1994).
125. Wilson, M. A. & McNaughton, B. L. Dynamics of the
hippocampal ensemble code for space. Science 261,
1055–1058 (1993).
Utilizing multiple recording tetrodes, these authors
monitored the simultaneous activity of more than
one-hundred CA1 neurons as a rat explored the
environment. This study was the first to look at
hippocampal spatial activity on the level of a network
and also the first to demonstrate the establishment of
a spatial code in a novel environment.
126. Eichenbaum, H., Dudchenko, P., Wood, W., Shapiro, M. &
Tanila, H. The hippocampus, memory, and place cells: is it
spatial memory or a memory space? Neuron 23, 209–226
(1999).
127. Lee, A. K. & Wilson, M. A. Memory of sequential experience
in the hippocampus during slow wave sleep. Neuron 36,
1183–1194 (2002).
128. Louie, K. & Wilson, M. A. Temporally structured replay of
awake hippocampal ensemble activity during rapid eye
movement sleep. Neuron 29, 145–156 (2001).
129. Poucet, B., Save, E. & Lenck-Santini, P. P. Sensory and
memory properties of hippocampal place cells. Rev.
Neurosci. 11, 95–111 (2000).
130. Moser, E. I. & Paulsen, O. New excitement in cognitive
space: between place cells and spatial memory. Curr. Opin.
Neurobiol. 11, 745–751 (2001).
The authors reviewed recent evidence connecting
hippocampal place cell activity with its mnemonic
function.
372 | MAY 2004 | VOLUME 5
131. Skaggs, W. E. & McNaughton, B. L. Replay of neuronal firing
sequences in rat hippocampus during sleep following spatial
experience. Science 271, 1870–1873 (1996).
132. Kudrimoti, H., Barnes, C. A. & McNaughton, B. L.
Reactivation of hippocampal cell assemblies: effects of
behavioral state, experience, and EEG dynamics.
J. Neurosci. 19, 4090–4101 (1999).
133. Nadasdy, Z., Hirase, H., Czurko, A., Csicsvari, J. & Buzsaki, G.
Replay and time compression of recurring spike sequences
in the hippocampus. J. Neurosci. 19, 9497–9507 (1999).
134. Muller, R. U. & Kubie, J. L. The effects of changes in the
environment on the spatial firing of hippocampal complex-
spike cells. J. Neurosci. 7, 1951–1968 (1987).
135. O’Keefe, J. & Conway, D. H. Hippocampal place units in the
freely moving rat: why they fire where they fire. Exp. Brain.
Res. 31, 573–590 (1978).
136. O’Keefe, J. & Speakman, A. Single unit activity in the rat
hippocampus during a spatial memory task. Exp. Brain Res.
68, 1–27 (1987).
137. Pico, R. M., Gerbrandt, L. K., Pondel, M. & Ivy, G. During
stepwise cue deletion, rat place behaviors correlate with
place unit responses. Brain Res. 330, 369–372 (1985).
138. Mehta, M. R., Barnes, C. A. & McNaughton, B. L. Experience-
dependent, asymmetric expansion of hippocampal place
fields. Proc. Natl Acad. Sci. USA 94, 8918–8921 (1997).
139. Mehta, M. R., Quirk, M. C. & Wilson, M. A. Experience-
dependent asymmetric shape of hippocampal receptive
fields. Neuron 25, 707–715 (2000).
140. Kentros, C. et al. Abolition of long-term stability of new
hippocampal place cell maps by NMDA receptor blockade.
Science 280, 2121–2126 (1998).
Combining pharmacology and in vivo recording, the
authors show that acute blockade of the NMDAR during
the exploration of a novel environment impairs the long-
term stability of the newly encoded place fields.
141. Ekstrom, A. D., Meltzer, J., McNaughton, B. L. & Barnes, C. A.
NMDA receptor antagonism blocks experience-dependent
expansion of hippcampal ‘place fields’. Neuron 31, 631–638
(2001).
The authors demonstrated, using the NMDAR
antagonist CPP, that experience-dependent
asymmetric place field expansion in hippocampal
CA1 is NR-dependent, but phase precession on the
theta wave is not.
142. Rotenberg, A., Mayford, M., Hawkins, R. D., Kandel, E. R. &
Muller, R. U. Mice expressing activated CaMKII lack low
frequency LTP and do not form stable place cells in the
CA1 region of the hippocampus. Cell 87, 1351–1361 (1996).
143. Rotenberg, A., Abel, T., Hawkins, R. D., Kandel, E. R. &
Muller, R. U. Parallel instabilities of long-term potentiation,
place cells, and learning caused by decreased protein
kinase A activity. J. Neurosci. 20, 8096–8102 (2000).
144. Yan, J. et al. Place-cell impairment in glutamate receptor 2
mutant mice. J. Neurosci. 22, RC204 (2002).
145. Cho, Y. H., Giese, K. P., Tanila, H., Silva, A. J. &
Eichenbaum, H. Abnormal hippocampal spatial
representations in aCaMKIIT286A and CREBα/– mice. Science
279, 867–869 (1998).
146. Muller, R. U., Bostock, E., Taube, J. S. & Kubie, J. L. On the
directional firing properties of hippocampal place cells.
J. Neurosci. 14, 7235–7251 (1994).
147. Anderson, M. I. & Jeffery, K. J. Heterogeneous modulation
of place cell firing by changes in context. J. Neurosci. 23,
8827–8835 (2003).
148. Wood, E. R., Dudchenko, P. A., Robitsek, R. J. &
Eichenbaum, H. Hippocampal neurons encode information
about different types of memory episodes occurring in the
same location. Neuron 27, 623–633 (2000).
149. Young, B. J., Fox, G. D. & Eichenbaum, H. Correlates of
hippocampal complex-spike cell activity in rats performing a
nonspatial radial maze task. J. Neurosci. 14, 6553–6563
(1994).
150. Hollup, S. A., Molden, S., Donnett, J. G., Moser, M. B. &
Moser, E. I. Accumulation of hippocampal place fields at the
goal location in an annular watermaze task. J. Neurosci. 21,
1635–1644 (2001).
©2004 Nature Publishing Group
151. Fyhn, M., Molden, S., Hollup, S., Moser, M.-B. & Moser, E. I.
Hippocampal neurons responding to first-time dislocation of
a target object. Neuron 35, 555–566 (2002).
152. Rolls, E. T. A theory of hippocampal function in memory.
Hippocampus 6, 601–620 (1996).
153. O’Keefe, J. Hippocampus, theta, and spatial memory. Curr.
Opin. Neurobiol. 3, 917–924. (1993).
154. O’Keefe, J. A review of the hippocampal place cells. Prog.
Neurobiol. 13, 419–349 (1979).
155. Muller, R. U., Stead, M. & Pach, J. The hippocampus as
a cognitive graph. J. Gen. Physiol. 107, 663–694
(1996).
156. Best, P. J., White, A. M. & Minai, A. Spatial processing in the
brain: the activity of hippocampal place cells. Annu. Rev.
Neurosci. 24, 459–486 (2001).
157. Thompson, L. T. & Best, P. J. Long-term stability of the
place-field activity of single units recorded from the dorsal
hippocampus of freely behaving rats. Brain Res. 509,
299–308 (1990).
158. Muller, R. A quarter of a century of place cells. Neuron 17,
813–822 (1996).
159. Redish, A. D. et al. Independence of firing correlates of
anatomically proximate hippocampal pyramidal cells.
J. Neurosci. 21, RC134 (2001).
160. Branda, C. S. & Dymecki, S. M. Talking about a revolution:
The impact of site-specific recombinases on genetic
analyses in mice. Dev. Cell 6, 7–28 (2004).
161. Morozov, A., Kellendonk, C., Simpson, E. & Tronche, F.
Using conditional mutagenesis to study the brain. Biol.
Psychiatry 54, 1125–1133 (2003).
162. Grant, S. G. N. et al. Impaired long-term potentiation, spatial
learning, and hippocampal development in fyn mutant mice.
Science 258, 1903–1909 (1992).
163. Blendy, J. A., Kaestner, K. H., Schmid, W., Gass, P. &
Schutz, G. Targeting of the CREB gene leads to up-
regulation of a novel CREB mRNA isoform. EMBO J. 15,
1098–1106 (1996).
164. Hummler, E. et al. Targeted mutation of the CREB gene:
compensation within the CREB/ATF family of transcription
factors. Proc. Natl Acad. Sci. USA 91, 5647–5651
(1994).
165. Mills, A. A. Changing colors in mice: an inducible system
that delivers. Genes Dev. 15, 1461–1467 (2001).
166. Schonig, K. & Bujard, H. Generating conditional mouse
mutants via tetracycline-controlled gene expression.
Methods Mol. Biol. 209, 69–104 (2003).
167. Gross, C. et al. Serotonin1A receptor acts during
development to establish normal anxiety-like behaviour in
the adult. Nature 416, 396–400 (2002).
168. Chen, J. et al. Transgenic animals with inducible, targeted
gene expression in brain. Mol. Pharmacol. 54, 495–503
(1998).
169. Mayford, M. et al. Control of memory formation through
regulated expression of a CaMKII transgene. Science 274,
1678–1683 (1996).
Acknowledgements
This work was supported by an individual NIH grant to S.T. and
an NIH Silvio O. Conte Center grant to S.T. and M.A.W., and by
grants from the Howard Hughes Medical Institute to S.T. and
from the RIKEN-MIT Neuroscience Research Center to S.T.
and M.A.W.
Competing interests statement
The authors declare that they have no competing financial interests.
Online links
FURTHER INFORMATION
Nakazawa’s homepage: http://neuroscience.nih.gov/
Lab.asp?Org_ID=496
Tonegawa’s homepage: http://www.hhmi.org/research
/investigators/tonegawa.html
Wilson’s homepage: http://web.mit.edu/bcs/people/wilson.shtml
Access to this interactive links box is free online.
www.nature.com/reviews/neuro